The Experts below are selected from a list of 234663 Experts worldwide ranked by ideXlab platform
Kai Ming Ting - One of the best experts on this subject based on the ideXlab platform.
-
Matching model versus single model: A study of the requirement to match Class Distribution using decision trees
Lecture Notes in Computer Science, 2004Co-Authors: Kai Ming TingAbstract:A tacit assumption in Classifier induction is that the Class Distribution of the training set must match the Class Distribution of the test set. A direct implementation is to retrain a model using a data set with matching Class Distribution every time the operating condition changes (i.e., the matching model). The alternative is to modify the decision rule of a previous trained model to the new operating condition. The latter is the single model approach commonly used and recommended by many researchers. In this paper, we argue with empirical support using decision trees that learning using the matching Class Distribution is desirable. We also make explicit the differences and limitations of the two methods for the single model approach: rescaling and thresholding.
-
ECML - Matching model versus single model: a study of the requirement to match Class Distribution using decision trees
Machine Learning: ECML 2004, 2004Co-Authors: Kai Ming TingAbstract:A tacit assumption in Classifier induction is that the Class Distribution of the training set must match the Class Distribution of the test set. A direct implementation is to retrain a model using a data set with matching Class Distribution every time the operating condition changes (i.e., the matching model). The alternative is to modify the decision rule of a previous trained model to the new operating condition. The latter is the single model approach commonly used and recommended by many researchers. In this paper, we argue with empirical support using decision trees that learning using the matching Class Distribution is desirable. We also make explicit the differences and limitations of the two methods for the single model approach: rescaling and thresholding.
-
Discovery Science - A Study on the Effect of Class Distribution Using Cost-Sensitive Learning
Discovery Science, 2002Co-Authors: Kai Ming TingAbstract:This paper investigates the effect of Class Distribution on the predictive performance of Classification models using cost-sensitive learning, rather than the sampling approach employed previously by a similar study. The predictive performance is measured using the cost space representation, which is a dual to the ROC representation. This study shows that Distributions which range between the natural Distribution and the balanced Distribution can also produce the best models, contrary to the finding of the previous study. In addition, we find that the best models are larger in size than those trained using the natural Distribution. We also show two different ways to achieve the same effect of the corrected probability estimates proposed by the previous study.
-
A study on the effect of Class Distribution using cost-sensitive learning
Lecture Notes in Computer Science, 2002Co-Authors: Kai Ming TingAbstract:This paper investigates the effect of Class Distribution on the predictive performance of Classification models using cost-sensitive learning, rather than the sampling approach employed previously by a similar study. The predictive performance is measured using the cost space representation, which is a dual to the ROC representation. This study shows that Distributions which range between the natural Distribution and the balanced Distribution can also produce the best models, contrary to the finding of the previous study. In addition, we find that the best models are larger in size than those trained using the natural Distribution. We also show two different ways to achieve the same effect of the corrected probability estimates proposed by the previous study.
Heli Peltola - One of the best experts on this subject based on the ideXlab platform.
-
effects of climate change and management on net climate impacts of production and utilization of energy biomass in norway spruce with stable age Class Distribution
Gcb Bioenergy, 2016Co-Authors: Piritta Torssonen, Antti Kilpelainen, Harri Strandman, Seppo Kellomaki, Kirsti Jylha, Antti Asikainen, Heli PeltolaAbstract:We studied the effects of climate change and forest management scenarios on net climate impacts (radiative forcing) of production and utilization of energy biomass, in a Norway spruce forest area over an 80-year simulation period in Finnish boreal conditions. A stable age-Class Distribution was used in model-based analyses to identify purely the management effects under the current and changing climate (SRES B1 and A2 scenarios). The radiative forcing was calculated based on an integrated use of forest ecosystem model simulations and a life cycle assessment (LCA) tool. In this work, forest-based energy was used to substitute coal, and current forest management (baseline management) was used as a reference management. In alternative management scenarios, the stocking was maintained 20% higher in thinning compared to the baseline management, and nitrogen fertilization was applied. Intensity of energy biomass harvest (e.g. logging residues, coarse roots and stumps) was varied in the final felling of the stands at the age of 80 years. Also, the economic profitability (NPV, 3% interest rate) of integrated production of timber and energy biomass was calculated for each management scenario. Our results showed that compared to the baseline management, climate benefits could be increased by maintaining higher stocking in thinning over rotation, using nitrogen fertilization and harvesting logging residues, stumps and coarse roots in the final felling. Under the gradually changing climate (in both SRES B1 and A2), the climate benefits were lower compared to the current climate. Trade-offs between NPV and net climate impacts also existed.
-
impacts of intensive management and landscape structure on timber and energy wood production and net co2 emissions from energy wood use of norway spruce
Bioenergy Research, 2012Co-Authors: Johanna Routa, Seppo Kellomaki, Heli PeltolaAbstract:The aim of this study was to analyze the effects of intensive management and forest landscape structure (in terms of age Class Distribution) on timber and energy wood production (m3 ha−1), net present value (NPV, € ha−1) with implications on net CO2 emissions (kg CO2 MWh−1 per energy unit) from energy wood use of Norway spruce grown on medium to fertile sites. This study employed simulations using a forest ecosystem model and the Emission Calculation Tool, considering in its analyses: timber (saw logs, pulp) and energy wood (small-sized stem wood and/or logging residuals for top part of stem, branches, and needles) from the first thinning and harvesting residuals and stumps from the final felling. At the stand level, both fertilization and high pre-commercial stand density clearly increased timber production and the amount of energy wood. Short rotation length (40 and 60 years) outputted, on average, the highest annual stem wood production (most fertile and medium fertile sites), the 60 year rotation also outputted the highest average annual net present value (NPV with interest rates of 1–4%). On the other hand, even longer rotation lengths, up to 80 and 100 years, were needed to output the lowest net CO2 emissions per year in energy wood use. At the landscape level, the largest productivity (both for timber and energy wood) was obtained using rotation lengths of 60 and 80 years with an initial forest landscape structure dominated by older mature stands (a right-skewed age-Class Distribution). If the rotation length was 120 years, the initial forest landscape dominated by young stands (a left-skewed age-Class Distribution) provided the highest productivity. However, the NPV with interest rate of 2% was, on average, the highest with a right-skewed Distribution regardless of the rotation length. If the rotation length was 120 years, normal age Class Distribution provided, on average, the highest NPV. On the other hand, the lowest emissions (kg CO2 MWh−1a−1) were obtained with the left-skewed age-Class Distribution using the rotation lengths of 60 and 80 years, and with the normal age-Class Distribution using the rotation length of 120 years. Altogether, the management regimes integrating both timber and energy wood production and using fertilization provided, on average, the lowest emissions over all management alternatives considered.
-
impacts of forest landscape structure and management on timber production and carbon stocks in the boreal forest ecosystem under changing climate
Forest Ecology and Management, 2007Co-Authors: Jordi Garciagonzalo, Heli Peltola, Ane Zubizarreta Gerendiain, Seppo KellomakiAbstract:Abstract A process-based model was used to assess the sensitivity of timber production and carbon (C) sequestration to the structure (in terms of age Class Distribution) of a boreal forest landscape and to the management under changing climatic conditions. Moreover, an approach to calculate the cost of C sequestration, through C sink enhancement, was used by computing the potential loss in the net present value (NPV) of timber when management with maximum C stock is chosen. Regardless of the climate scenario and initial age Class Distribution used, management had a clear effect on the mean C stock in the forest ecosystem. Any management regime allowing a higher tree stocking than business-as-usual management increased the timber production and simultaneously maintained or increased the C stock in the forest ecosystem. On the other hand, the maximum C stock in the forest and the lowest NPV were observed when no thinning was applied before the final cut. The changing climate increased forest productivity and also C stock. When using the same management for the entire management unit, the initial age Class Distribution had a large influence on the results of timber production (up to 20% difference) but not on C stock in the forest ecosystem (3%).
Seppo Kellomaki - One of the best experts on this subject based on the ideXlab platform.
-
effects of climate change and management on net climate impacts of production and utilization of energy biomass in norway spruce with stable age Class Distribution
Gcb Bioenergy, 2016Co-Authors: Piritta Torssonen, Antti Kilpelainen, Harri Strandman, Seppo Kellomaki, Kirsti Jylha, Antti Asikainen, Heli PeltolaAbstract:We studied the effects of climate change and forest management scenarios on net climate impacts (radiative forcing) of production and utilization of energy biomass, in a Norway spruce forest area over an 80-year simulation period in Finnish boreal conditions. A stable age-Class Distribution was used in model-based analyses to identify purely the management effects under the current and changing climate (SRES B1 and A2 scenarios). The radiative forcing was calculated based on an integrated use of forest ecosystem model simulations and a life cycle assessment (LCA) tool. In this work, forest-based energy was used to substitute coal, and current forest management (baseline management) was used as a reference management. In alternative management scenarios, the stocking was maintained 20% higher in thinning compared to the baseline management, and nitrogen fertilization was applied. Intensity of energy biomass harvest (e.g. logging residues, coarse roots and stumps) was varied in the final felling of the stands at the age of 80 years. Also, the economic profitability (NPV, 3% interest rate) of integrated production of timber and energy biomass was calculated for each management scenario. Our results showed that compared to the baseline management, climate benefits could be increased by maintaining higher stocking in thinning over rotation, using nitrogen fertilization and harvesting logging residues, stumps and coarse roots in the final felling. Under the gradually changing climate (in both SRES B1 and A2), the climate benefits were lower compared to the current climate. Trade-offs between NPV and net climate impacts also existed.
-
impacts of intensive management and landscape structure on timber and energy wood production and net co2 emissions from energy wood use of norway spruce
Bioenergy Research, 2012Co-Authors: Johanna Routa, Seppo Kellomaki, Heli PeltolaAbstract:The aim of this study was to analyze the effects of intensive management and forest landscape structure (in terms of age Class Distribution) on timber and energy wood production (m3 ha−1), net present value (NPV, € ha−1) with implications on net CO2 emissions (kg CO2 MWh−1 per energy unit) from energy wood use of Norway spruce grown on medium to fertile sites. This study employed simulations using a forest ecosystem model and the Emission Calculation Tool, considering in its analyses: timber (saw logs, pulp) and energy wood (small-sized stem wood and/or logging residuals for top part of stem, branches, and needles) from the first thinning and harvesting residuals and stumps from the final felling. At the stand level, both fertilization and high pre-commercial stand density clearly increased timber production and the amount of energy wood. Short rotation length (40 and 60 years) outputted, on average, the highest annual stem wood production (most fertile and medium fertile sites), the 60 year rotation also outputted the highest average annual net present value (NPV with interest rates of 1–4%). On the other hand, even longer rotation lengths, up to 80 and 100 years, were needed to output the lowest net CO2 emissions per year in energy wood use. At the landscape level, the largest productivity (both for timber and energy wood) was obtained using rotation lengths of 60 and 80 years with an initial forest landscape structure dominated by older mature stands (a right-skewed age-Class Distribution). If the rotation length was 120 years, the initial forest landscape dominated by young stands (a left-skewed age-Class Distribution) provided the highest productivity. However, the NPV with interest rate of 2% was, on average, the highest with a right-skewed Distribution regardless of the rotation length. If the rotation length was 120 years, normal age Class Distribution provided, on average, the highest NPV. On the other hand, the lowest emissions (kg CO2 MWh−1a−1) were obtained with the left-skewed age-Class Distribution using the rotation lengths of 60 and 80 years, and with the normal age-Class Distribution using the rotation length of 120 years. Altogether, the management regimes integrating both timber and energy wood production and using fertilization provided, on average, the lowest emissions over all management alternatives considered.
-
impacts of forest landscape structure and management on timber production and carbon stocks in the boreal forest ecosystem under changing climate
Forest Ecology and Management, 2007Co-Authors: Jordi Garciagonzalo, Heli Peltola, Ane Zubizarreta Gerendiain, Seppo KellomakiAbstract:Abstract A process-based model was used to assess the sensitivity of timber production and carbon (C) sequestration to the structure (in terms of age Class Distribution) of a boreal forest landscape and to the management under changing climatic conditions. Moreover, an approach to calculate the cost of C sequestration, through C sink enhancement, was used by computing the potential loss in the net present value (NPV) of timber when management with maximum C stock is chosen. Regardless of the climate scenario and initial age Class Distribution used, management had a clear effect on the mean C stock in the forest ecosystem. Any management regime allowing a higher tree stocking than business-as-usual management increased the timber production and simultaneously maintained or increased the C stock in the forest ecosystem. On the other hand, the maximum C stock in the forest and the lowest NPV were observed when no thinning was applied before the final cut. The changing climate increased forest productivity and also C stock. When using the same management for the entire management unit, the initial age Class Distribution had a large influence on the results of timber production (up to 20% difference) but not on C stock in the forest ecosystem (3%).
Andreas Brune - One of the best experts on this subject based on the ideXlab platform.
-
Dynamics in composition and size-Class Distribution of humic substances in profundal sediments of Lake Constance
Organic Geochemistry, 2001Co-Authors: Andreas Kappler, Bernhard Schink, Andreas BruneAbstract:Organic matter from profundal lake sediments was fractionated and analyzed with high depth resolution, and the size-Class Distribution of humic substances was determined by gel-permeation chromatography. Together with a depthdependent increase of the thermal stability of the organic matter and an increasing C/N atomic ratio of the humic acids, the size-Distribution patterns of diAerent fractions indicated an increasing degree of humification with depth. The main stabilization processes occurred in the surface layer (0‐2 cm depth), which is clearly distinguished from the deeper sediment. The depth profiles also indicated that humification in lake sediments consists of simultaneous degradation and polymerization processes. # 2001 Elsevier Science Ltd. All rights reserved.
-
influence of gut alkalinity and oxygen status on mobilization and size Class Distribution of humic acids in the hindgut of soil feeding termites
Applied Soil Ecology, 1999Co-Authors: Andreas Kappler, Andreas BruneAbstract:The majority of termite species do not feed on wood, but have a humivorous mode of nutrition. However, the exact nature of the carbon and energy source of soil-feeding termites, and also the role of the extreme alkalinity in their anterior hindguts are still unknown. Using soil-feeding Cubitermes species and high-pressure gel-permeation chromatography, we found a significant shift toward lower molecular weight in the humic acid samples extracted from different hindgut regions, compared to the parent soil ingested by the termites. The changes in size-Class Distribution were most pronounced in the highly alkaline (pH 12) P1 compartment, where the apparent molecular weight calculated for the peak maximum of the most abundant medium-molecular-weight fraction decreased by30%. Microelectrode profiles demonstrated steep oxygen gradients into the alkaline hindgut compartments, caused by intestinal oxygen consumption in the gut periphery. In the less alkaline P3 region, oxygen consumption was at least partially attributable to biological processes, whereas in the highly alkaline P1 compartment, it seemed to be largely due to a chemical process. In vitro extraction of parent soil with NaOH solutions of increasing concentration confirmed that extraction efficiency of humic acids was not only enhanced by a high pH, but also by the simultaneous presence of oxygen. Similar results were obtained with 200 mM potassium carbonate solution, which mimics the pH and potassium concentration in the P1 region. However, in these experiments the pronounced molecular-weight shift found in the alkaline hindgut compartments was never observed, which indicates that additional factors must be present within the gut. We conclude that the extreme alkalinity in the anterior hindgut, supported by autoxidative processes, facilitates not only desorption of humic substances from the mineral matrix, but also decreases their molecular weight and increases their solubility. This renders so far unknown constituents of the humic matter accessible to microbial degradation in the subsequent, less alkaline hindgut compartments. # 1999 Elsevier Science B.V. All rights reserved.
Andreas Kappler - One of the best experts on this subject based on the ideXlab platform.
-
Dynamics in composition and size-Class Distribution of humic substances in profundal sediments of Lake Constance
Organic Geochemistry, 2001Co-Authors: Andreas Kappler, Bernhard Schink, Andreas BruneAbstract:Organic matter from profundal lake sediments was fractionated and analyzed with high depth resolution, and the size-Class Distribution of humic substances was determined by gel-permeation chromatography. Together with a depthdependent increase of the thermal stability of the organic matter and an increasing C/N atomic ratio of the humic acids, the size-Distribution patterns of diAerent fractions indicated an increasing degree of humification with depth. The main stabilization processes occurred in the surface layer (0‐2 cm depth), which is clearly distinguished from the deeper sediment. The depth profiles also indicated that humification in lake sediments consists of simultaneous degradation and polymerization processes. # 2001 Elsevier Science Ltd. All rights reserved.
-
influence of gut alkalinity and oxygen status on mobilization and size Class Distribution of humic acids in the hindgut of soil feeding termites
Applied Soil Ecology, 1999Co-Authors: Andreas Kappler, Andreas BruneAbstract:The majority of termite species do not feed on wood, but have a humivorous mode of nutrition. However, the exact nature of the carbon and energy source of soil-feeding termites, and also the role of the extreme alkalinity in their anterior hindguts are still unknown. Using soil-feeding Cubitermes species and high-pressure gel-permeation chromatography, we found a significant shift toward lower molecular weight in the humic acid samples extracted from different hindgut regions, compared to the parent soil ingested by the termites. The changes in size-Class Distribution were most pronounced in the highly alkaline (pH 12) P1 compartment, where the apparent molecular weight calculated for the peak maximum of the most abundant medium-molecular-weight fraction decreased by30%. Microelectrode profiles demonstrated steep oxygen gradients into the alkaline hindgut compartments, caused by intestinal oxygen consumption in the gut periphery. In the less alkaline P3 region, oxygen consumption was at least partially attributable to biological processes, whereas in the highly alkaline P1 compartment, it seemed to be largely due to a chemical process. In vitro extraction of parent soil with NaOH solutions of increasing concentration confirmed that extraction efficiency of humic acids was not only enhanced by a high pH, but also by the simultaneous presence of oxygen. Similar results were obtained with 200 mM potassium carbonate solution, which mimics the pH and potassium concentration in the P1 region. However, in these experiments the pronounced molecular-weight shift found in the alkaline hindgut compartments was never observed, which indicates that additional factors must be present within the gut. We conclude that the extreme alkalinity in the anterior hindgut, supported by autoxidative processes, facilitates not only desorption of humic substances from the mineral matrix, but also decreases their molecular weight and increases their solubility. This renders so far unknown constituents of the humic matter accessible to microbial degradation in the subsequent, less alkaline hindgut compartments. # 1999 Elsevier Science B.V. All rights reserved.